Equalizing loudspeakers based on anechoic measurements (community project)

[TimVG]

Hi,

Let's start with the F206 only. No EQ: 6.27
It is not easiest to EQ but we'll see how it pans out.

The score are probably not that accurate because they are not from the raw data.
Take them as relative scores rather.

View attachment 75811

again, can you discriminate between them and the raw speaker?
If so is there one better than the others?
I did not see, maybe I missed it, any clear answer for the M105, could you give us your final conclusion at one point?

% H1: 7.03
x0=[ 221.0, -0.66,  6.43,...
     375.0,  0.95,  2.53,...
     598.0,  0.48,  6.50,...
    1100.0, -0.33,  0.50,...
    1324.0,  1.65,  3.82,...
    2145.0,  0.73,  5.30,...
    3435.0,  1.13,  3.00,...
    5340.0,  0.75,  8.40,...
   10010.0, -0.50,  11.1,...
   11560.0, -0.75,  3.80,...
    4333.0, -0.00,  3.60,...
    1];

% SL: 6.94
x0=[ 223.0, -0.46,  9.00,...
     375.0,  1.40,  2.53,...
     570.0,  0.51,  6.50,...
     987.0, -0.43,  0.61,...
    1321.0,  2.00,  4.00,...
    1820.0, -1.24,  3.70,...
    3435.0,  1.65,  3.00,...
    5350.0,  0.75,  12.5,...
    9924.0, -0.67,  7.60,...
   11560.0, -1.10,  2.60,...
    4333.0, -0.00,  3.60,...
    1];

% SC 7.14
x0=[ 223.0, -0.46,  11.1,...
     375.0,  1.40,  2.53,...
     570.0,  0.66,  4.70,...
     987.0, -0.33,  0.69,...
    1331.0,  2.00,  3.22,...
    1721.0, -0.88,  5.45,...
    3435.0,  0.90,  6.00,...
    5351.0,  0.69,  14.3,...
    9924.0, -0.61,  8.60,...
   11560.0, -0.95,  2.00,...
    4333.0, -0.00,  3.60,...
    1];

So, this took longer than expected. I can't say the differences were dramatic with any of them compared to the original. I decided first to find out which filter I liked best on a variety of material, and then compare those to the original and my own. After testing blind over the course of two days, I settled on filter what turned out to be filter #2. I found it overall less fatiguing than the other two.

Today I performed a new blind test compared to the original and my own filterset, today - I really had to split this over some days as it does become tiring and my results seem to become less reliable as time passes.

I ended up preferring my own filter as the best, followed by your #2 filter, followed by the uncorrected speaker. I found my filter sounded the smoothest and most balanced on most of the material. The big difference I found was overall balance, with yours sounding a bit brighter/clearer.

Now, I haven't loaded yours into REW yet, so I'll let you do the honors and if you care to calculate the score to my filters.. I'd be curious to know - here are my filters:

 

[Robbo99999]

So, this took longer than expected. I can't say the differences were dramatic with any of them compared to the original. I decided first to find out which filter I liked best on a variety of material, and then compare those to the original and my own. After testing blind over the course of two days, I settled on filter what turned out to be filter #2. I found it overall less fatiguing than the other two.

Today I performed a new blind test compared to the original and my own filterset, today - I really had to split this over some days as it does become tiring and my results seem to become less reliable as time passes.

I ended up preferring my own filter as the best, followed by your #2 filter, followed by the uncorrected speaker. I found my filter sounded the smoothest and most balanced on most of the material. The big difference I found was overall balance, with yours sounding a bit brighter/clearer.

Now, I haven't loaded yours into REW yet, so I'll let you do the honors and if you care to calculate the score to my filters.. I'd be curious to know - here are my filters:

View attachment 76159

Not massive differences in preference then between those 2 EQ strategies then given that your preference for each of them swapped places on the two different days.....is my initial impression.

Kudos to you for blind testing like this, because I imagine it's not easy to do & it's rare!

 

[TimVG]

Not massive differences in preference then between those 2 EQ strategies then given that your preference for each of them swapped places on the two different days.....is my initial impression.

Kudos to you for blind testing like this, because I imagine it's not easy to do & it's rare!

Well not really swapped places as the initial two days was Maiky's 3 filters against one another - I settled on the one that I liked the best over the course of those two days. Some tracks showed the differences more than others, so I could have just limited myself there - but to me I like to keep a broad perspective on these things.

Then today I tested the #2 filter, against the uncorrected F206 and my own custom filter - which in the end I still preferred, even when not knowing what was playing. I did this because switching between 5 EQ filters is just insane when doing it blind - especially when they are not that different from one another. Maybe others could pull it off, but my results would be all over the place.

It takes some prep as my girlfriend has to rename the files, and I have to make sure I don't see any EQ curves in the process - but I can't say I'm not enjoying it - although after a good hour or so I really need to go do something else to keep it so.

 

[Robbo99999]

Well not really swapped places as the initial two days was Maiky's 3 filters against one another - I settled on the one that I liked the best over the course of those two days. Some tracks showed the differences more than others, so I could have just limited myself there - but to me I like to keep a broad perspective on these things.

Then today I tested the #2 filter, against the uncorrected F206 and my own custom filter - which in the end I still preferred, even when not knowing what was playing. I did this because switching between 5 EQ filters is just insane when doing it blind - especially when they are not that different from one another. Maybe others could pull it off, but my results would be all over the place.

It takes some prep as my girlfriend has to rename the files, and I have to make sure I don't see any EQ curves in the process - but I can't say I'm not enjoying it - although after a good hour or so I really need to go do something else to keep it so.

Your girlfriend is very......dedicated! Yep, I see what you mean now with your distinction you're making in this post, I didn't see that distinction in your previous post.....ok, yeah, I think that's a valid preference then for your own filter, just seeing it based on what you describe.

 

[QMuse]

@edechamps, you may find this interesting. This is how PIRDI looks with my right speaker, to the best of my limited ability to measure it in room.

Upper curve (light green) is ON response measured from app 40cm distance. Lower curve is response from LP measured with MMM from 4 meters from the speaker. Both curves are smoothed wth 1/6. XO is LR24 at 1800Hz.

Blue curve is PIRDI. My guess is that 500-800Hz region is lower with ON curve because mic didn't catch it from such close distance. My speakers are towers, one woofer is firing upwards from the top of the speaker and the other is firing front sitting on top of the twitter so probably a little more distance than 40cm is needed to catch 500-800Hz region correctly.

Anyway, this looks pretty good to me, it also sounds pretty good.

This is IR from the ON measurement to check the ammount of reflections that interferred with the direct sound. I think they are reasonably small.

This is the filter I aplied in that region:

 

[QMuse]

This does lend credence to my idea that the Olive model selected PIR not because a good in-room response causes good sound, but because a well-behaved PIR correlates with a better-sounding speaker in general.

I really don't see it that way. PIR is a composite response derived from anechoic measurements. It is a result of speaker design and drivers that were used and as such PIR represents a design target, at least if you want the speaker to sound good to listeners and to achieve a good score. If you choose to use EQ as part of the speaker design the EQ filter becomes part of the design process in order to achieve good sound/good score and smooth PIR will be the measured result of such design effort. No response curve is causing the good sound, speaker is doing that via it's box, drivers and EQ. Response curve is only a representative of the sound and as such cannot cause it. But EQ which results in a smooth PIR can contribute to a good sound as much as high quality drivers and carefully designed box.

 

[JIM_82]

On "classic" measurement of speaker or horn directivity, response is normalized on axis.
This is not necessarily done on a spinorama, especially for speakers not equipped with DSP.

This normalization is the ideal EQ of the speaker.
And as Toole says, if the other curves are not good, no EQ will correct, return to design.

The true difficulty of EQ in room is under schroeder frequency.

 

[Maiky76]

Hi,

So, this took longer than expected. I can't say the differences were dramatic with any of them compared to the original. I decided first to find out which filter I liked best on a variety of material, and then compare those to the original and my own. After testing blind over the course of two days, I settled on filter what turned out to be filter #2. I found it overall less fatiguing than the other two.

Today I performed a new blind test compared to the original and my own filterset, today - I really had to split this over some days as it does become tiring and my results seem to become less reliable as time passes.

I ended up preferring my own filter as the best, followed by your #2 filter, followed by the uncorrected speaker. I found my filter sounded the smoothest and most balanced on most of the material. The big difference I found was overall balance, with yours sounding a bit brighter/clearer.

Now, I haven't loaded yours into REW yet, so I'll let you do the honors and if you care to calculate the score to my filters.. I'd be curious to know - here are my filters:

View attachment 76159

Hi,

Kudos for going though the listening tests. These can indeed be very tiring.
Your EQ score is 6.87, again it more of an indicative score than a real value.
Especially since the data measured on a different speaker.

From the curves it looks like you have EQed for a smooth PIR rather than ON/LW, is that correct?
The decrease at HF you came up with would be typical of what i saw optimizing the based on the score with a speaker with constanat directivity waveguide on the TW.
Was your EQ adjusted by ear?

M

 

[TimVG]

Hi,


Hi,

Kudos for going though the listening tests. These can indeed be very tiring.
Your EQ score is 6.87, again it more of an indicative score than a real value.
Especially since the data measured on a different speaker.
View attachment 76387
From the curves it looks like you have EQed for a smooth PIR rather than ON/LW, is that correct?
The decrease at HF you came up with would be typical of what i saw optimizing the based on the score with a speaker with constanat directivity waveguide on the TW.
Was your EQ adjusted by ear?
View attachment 76388

M

Hi, thanks for the graphs and the continued effort on your part as well.

My own EQ came in 3 steps.
-First is EQ'ing the LW to be moreor less flat. I pay less attention >10kHz since Harman graphs seem to always show a little bit more here than Amir or others measure.

-Second is using these corrections looking at the on axis curve and subsequently tilting it down if needed to be sure there is no upward trend. No PEQ filters here, just a shelf.

-third I will look at the PIR if there are further possible anomalies. Any corrections here are kept small in any case.

If you'd like we can do the C208 soon, although I'm currently on a short vaction and will be back wednesday.

 

[spacevector]

Hi folks - have EQ filters been created for the SVS Ultra bookshelf speaker yet? I am hoping Amir will try to listen with these applied.

 

[Maiky76]

Hi,

So, this took longer than expected. I can't say the differences were dramatic with any of them compared to the original. I decided first to find out which filter I liked best on a variety of material, and then compare those to the original and my own. After testing blind over the course of two days, I settled on filter what turned out to be filter #2. I found it overall less fatiguing than the other two.

Today I performed a new blind test compared to the original and my own filterset, today - I really had to split this over some days as it does become tiring and my results seem to become less reliable as time passes.

I ended up preferring my own filter as the best, followed by your #2 filter, followed by the uncorrected speaker. I found my filter sounded the smoothest and most balanced on most of the material. The big difference I found was overall balance, with yours sounding a bit brighter/clearer.

Now, I haven't loaded yours into REW yet, so I'll let you do the honors and if you care to calculate the score to my filters.. I'd be curious to know - here are my filters:

View attachment 76159

Hi,

Kudos for going though the listening tests. These can indeed be very tiring.
Your EQ score is 6.87, again it more of an indicative score than a real value.
View attachment 76387
From the curves it looks like you have EQed for a smooth PIR rather than ON/LW, is that correct?
Was your EQ adjusted by ear?
View attachment 76388

M

Hi, thanks for the graphs and the continued effort on your part as well.

My own EQ came in 3 steps.
-First is EQ'ing the LW to be moreor less flat. I pay less attention >10kHz since Harman graphs seem to always show a little bit more here than Amir or others measure.

-Second is using these corrections looking at the on axis curve and subsequently tilting it down if needed to be sure there is no upward trend. No PEQ filters here, just a shelf.

-third I will look at the PIR if there are further possible anomalies. Any corrections here are kept small in any case.

If you'd like we can do the C208 soon, although I'm currently on a short vaction and will be back wednesday.

There you go:
All PEQ except the one LowShelf (LS) per filter.
You might want to adjust the volume as the EQed response are lower than the raw speaker.

% 6.54 MX
01=[  50.0,  1.42,  1.80,...
     613.0, -0.86,  2.43,...
    2050.0, -1.00,  1.92,...
    1045.0, -0.67,  0.70,... LS
    5740.0, -2.62,  1.71,...
   11045.0, -1.45,  2.81,...

%  6.58 HM
02=[  50.0,  1.42,  1.80,...
     659.0, -0.86,  2.00,...
    1590.0, -1.00,  1.34,...
     932.0, -0.67,  0.70,... LS
    5740.0, -2.35,  1.50,...
   10950.0, -1.62,  2.41,...

%  6.57 HO
03=[  50.0,  1.42,  1.80,...
     377.0,  0.65,  2.57,...
     640.0, -0.61,  2.00,...
    1592.0, -0.70,  1.34,...
    1225.0, -0.60,  0.70,... LS
    5740.0, -2.15,  2.00,...
   10580.0, -1.10,  3.41,...

 

[QMuse]

Hi folks - have EQ filters been created for the SVS Ultra bookshelf speaker yet? I am hoping Amir will try to listen with these applied.

This speaker doesn't have any obvious flaws that need to be EQ-ed. It's spinorama charts are looking very good and the same is reflected with the high Olive's score. Speakers like this one, Revel M16, Kef R3 etc only need to be EQ-ed to your specific listening room/position in the range below 400Hz (or so).

 

[spacevector]

This speaker doesn't have any obvious flaws that need to be EQ-ed. It's spinorama charts are looking very good and the same is reflected with the high Olive's score. Speakers like this one, Revel M16, Kef R3 etc only need to be EQ-ed to your specific listening room/position in the range below 400Hz (or so).

Thoughts on this one?
https://www.audiosciencereview.com/...helf-speaker-review.15055/page-11#post-474234

 

[spacevector]

Hi folks who are creating this EQ:

First of all thank you for doing the work and making it available freely. I appreciate the effort and good will.

I have a question about the filter sets being created. It looks like these are generated programatically to meet a certain goal (say maximization of Olive score). The question/suggestion I have is: whether these can be simplified to some extent? I have pasted below, as an example, the EQ for JBL Studio 530. Total 20 filters are recommended. This seems like a lot. Can the filters be simplified a bit to reduce the number of PEQ's while still maintaining resultant SQ?

https://github.com/pierreaubert/spinorama/blob/develop/datas/eq/JBL Studio 530/iir.txt

Equaliser: Generic
JBL Studio 530 PIR.txt
Filter  1: ON  PK       Fc     345 Hz  Gain   1.2 dB  Q 10.000
Filter  2: ON  PK       Fc     382 Hz  Gain  -1.1 dB  Q 15.000
Filter  3: ON  PK       Fc     736 Hz  Gain  -1.5 dB  Q 1.000
Filter  4: ON  PK       Fc     932 Hz  Gain  -0.6 dB  Q 10.000
Filter  5: ON  PK       Fc    1170 Hz  Gain   0.8 dB  Q 8.000
Filter  6: ON  PK       Fc    1350 Hz  Gain  -0.6 dB  Q 16.000
Filter  7: ON  PK       Fc    1414 Hz  Gain  -1.7 dB  Q 5.000
Filter  8: ON  PK       Fc    1643 Hz  Gain -11.0 dB  Q 5.000
Filter  9: ON  PK       Fc    1648 Hz  Gain  12.0 dB  Q 4.998
Filter 10: ON  PK       Fc    1940 Hz  Gain   0.7 dB  Q 22.000
Filter 11: ON  PK       Fc    2114 Hz  Gain  -1.7 dB  Q 5.000
Filter 12: ON  PK       Fc    2220 Hz  Gain   0.6 dB  Q 25.000
Filter 13: ON  PK       Fc    2351 Hz  Gain  -1.2 dB  Q 4.997
Filter 14: ON  PK       Fc    3486 Hz  Gain  -1.4 dB  Q 3.773
Filter 15: ON  PK       Fc    5487 Hz  Gain  -2.6 dB  Q 1.001
Filter 16: ON  PK       Fc    8743 Hz  Gain  -3.3 dB  Q 5.000
Filter 17: ON  PK       Fc    9460 Hz  Gain  -0.8 dB  Q 22.000
Filter 18: ON  PK       Fc   10275 Hz  Gain   1.5 dB  Q 5.000
Filter 19: ON  PK       Fc   11146 Hz  Gain  -2.1 dB  Q 5.000
Filter 20: ON  PK       Fc   12588 Hz  Gain  -1.5 dB  Q 5.000

 

[flipflop]

It looks like these are generated programatically to meet a certain goal (say maximization of Olive score).

That one was made to have a PIR with a -1.0 dB/oct. slope. You can see the effects of the EQ by going here and choosing 'With EQ' or 'Ref. v.s. EQ'.

This seems like a lot. Can the filters be simplified a bit to reduce the number of PEQ's while still maintaining resultant SQ?

You should be able to remove filter 4, 6, 10, 12, and 17 without any real changes to the sound quality.

 

[spacevector]

That one was made to have a PIR with a -1.0 dB/oct. slope. You can see the effects of the EQ by going here and choosing 'With EQ' or 'Ref. v.s. EQ'.

You should be able to remove filter 4, 6, 10, 12, and 17 without any real changes to the sound quality.

Way cool. Thanks!

Can future EQ's be made to simplified version?

Are you using a script to generate these?

 

[flipflop]

Can future EQ's be made to simplified version?

I'll check with pierre if we can host multiple EQs for the same speaker like oratory1990 does with headphones. Then we could have a 10 band EQ for each speaker.

Are you using a script to generate these?

I use the 'Match response to target' function in REW, but it only gets the job halfway done, so there's still a lot of manual work that went into making these EQs.

 

[Pio2001]

Every instance of improving a loudspeaker through correcting the estimated in-room response have also shown an improvement in the direct sound

That is certainly not true for Neumann KH-120 speakers : https://www.audiosciencereview.com/...asurements-take-two.11323/page-22#post-425238

Or more generally for Genelec or Neumann speakers. Since their direct sound is dead flat, any modification will harm it.

I have compared in-room equalization vs flat on-axis response with the KH-120. I have always preferred the flat on axis response, but not for the reason discussed in this topic. What I didn't like with in-room equalization was the overall slope of the resulting curve.

I have never managed to get an in-room eq that have exactly the same overall slope as no eq / flat on-axis frequency response.
I think that this overall slope, that I can't get right, bothers me more, when I'm listening to the result, than the small variations in the PIRDI.

 

[ernestcarl]

The back and forth discussion in this thread is pretty interesting. Some of which is way over my head.

I've had to cobble my own EQ strategy for the Sceptre S8 monitors as I don't exactly have spinorama and directivity data for them. My particular room and placement is also not in the best way situated for taking multiple measurements to generate clean directivity plots.

Also, these monitors are not as flat out-of-the-box as one would hope given that these already use numerous filters internally with its own DSP.

While I mainly use MMM RTA+Peak measurements to generate my own PEQs at the MLP (2m distance) and elsewhere, not having reflection-free anechoic & spinorama data made me think that maybe I could do just a little better, yet again... hay... it's turning into an obsession now.

The Sceptre S8s are a coaxial design and use a horn waveguide to control beamwidth and directivity, maybe flattening the extreme nearfield (20cm or closer) on-axis response would produce some 'finer' EQ benefits not possible with my measurements taken at a distance?

So I did just that:

Forgive the crowded mess below:

*As a few reviewers have pointed out, these monitors' particular voicing emphasizes upper mids and treble more, leading most people to suggest using the monitor's internal HF shelving.

First curves are Sub+L&R log sweeps at MLP (2m distance), 1/6 smoothing, post EQ.

Second set of curves are MMM peak and RTA curves respectively, 1/6 smooothing, post EQ.

Third one is L&R only with NO EQ, nada! Just as they are at measured exactly 20cm from the opening of the horn throat.

Last one is with ALL my MMM RTA+Peak constructed PEQs -- at the MLP, and extreme sides of couch (very much taken into consideration) -- as well as with a few additional small adjustments and extra PEQs to flatten more using the measurements taken at extreme nearfield.

I made a few adjustements here and there -- until I got tired of it -- but left the larger majority of my former set of PEQs intact.

If you have the Sceptre S8, you can probably use the same common set of PEQs listed above, and get reasonably good results - a few manual adjustments and additional filters on your end is expected esp. below 500Hz!

Here's the full set as it currently stands:

Not too easy to A/B filter sets as almost half of them are within JRiver (not enough slots in the miniDSP in this case). I assume most of the important ones are in the miniDSP and the rest of the fluff in JRiver. LOL

I often am using miniDSP to play from external sources other than my HTPC, so it's necessary to split things apart for now.

I'd say over 95% of my current EQ was made with in-room measurements at a distance, and the last 5% with minor modifications to the original from the nearfield, on-axis graphs.

So does the final result hold up not only on-axis but off-axis as well?

Yep...

The far off-center seating area curves look somewhat worse, but I assure you, it still sounds great to me.

The wider listening area for context:

If there's only one person listening (nearly always), one can apply the simplest volume and delay adjustments to get the stereo image balance at center position again:

Evidently, as you move further off-axis, the response gets worse. A few PEQs in addition to basic volume level and time delay adjustments will easily flatten the response (much closer resembling the original on-axis curves) and give one a more accurate stereo image when off-axis, if so desired. The sound is already good enough as it is, though.

By the way, at the extreme ends of the couch, I captured way more reflections than direct sound itself. Basically, I point the microphone almost parallel to the side walls, and sometimes a little forward toward the speaker being measured. It's all being averaged out (well over 100 data points with each MMM RTA+Peak curves generated).

This is how they appear together:

And 1/6 smoothing

My older EQ preset results befor taking the nearfield measurements into consideration look a tad better in some places:

*Maybe looks a little smoother here for the far off-center seats as I used to point the mic more towards the speaker in the above, and less so directluy against the wall.

Again, since there is no easy way to A/B all filter sets near instantaneously, it's hard to say which I like better. I'd say I'm happy with both. In fact, I was already happy without any EQ above the transition zone in the past already. So I am not the best judge when it comes sound quality with pre-post EQ'ing speakers... so as long as said speakers already sound great out-of-the-box in the first place!

P.S. the treble is a bit more shelved, and sounds shelved, in the former set of preset and actual measurements. Tastes vary so how much of that you want will vary from person to person -- and even room to room.

 

[TimVG]

That is certainly not true for Neumann KH-120 speakers : https://www.audiosciencereview.com/...asurements-take-two.11323/page-22#post-425238

Or more generally for Genelec or Neumann speakers. Since their direct sound is dead flat, any modification will harm it.

I have compared in-room equalization vs flat on-axis response with the KH-120. I have always preferred the flat on axis response, but not for the reason discussed in this topic. What I didn't like with in-room equalization was the overall slope of the resulting curve.

I have never managed to get an in-room eq that have exactly the same overall slope as no eq / flat on-axis frequency response.
I think that this overall slope, that I can't get right, bothers me more, when I'm listening to the result, than the small variations in the PIRDI.

Yes, I agree. I mean that every time where someone corrected the PIR with audible succes it just happened to (by chance due to directivity) also in that case improve the direct sound.